GenAI Explainability
GenAI Explainability delivers plain-language explanations for RiskOS™ risk scores and identity decisions within a secure environment.
Behind every risk score and identity match is a team of fraud, compliance, and support professionals working to protect their organizations and legitimate users. When AI decisions are a black box, those teams spend hours decoding model logic, escalating to data science, and second-guessing outcomes. GenAI Explainability changes that.
GenAI Explainability transforms dense risk signals into plain-language summaries, so analysts understand why a decision was made—tailored to the specific case—and resolve cases in minutes instead of hours. It brings natural-language explanations to the key outputs across Socure's platform, surfacing the "why" behind risk scores, identity match decisions, and document verification results. Each explanation is based on the specific data from that evaluation and tailored to its unique signals. Once generated, an explanation is saved to the evaluation and isn't regenerated, ensuring consistency and auditability.
Unlike off-the-shelf AI tools, GenAI Explainability runs entirely within Socure's secure, sovereign environment. That means:
- No data leaves Socure's platform
- No LLM training leakage
- Full control, privacy, and compliance at every step
This capability does not affect decision outcomes. Instead, it makes them more transparent, understandable, and actionable for risk and compliance teams.
Supported products
GenAI Explainability is available for the following products:
- Verify (KYC)
- Verify+
- Predictive DocV
- Sigma Identity Fraud
- Sigma Synthetic Fraud
- Email Risk
- Phone Risk
- Address Risk
- Watchlist
Access GenAI Explainability
Socure enables GenAI Explainability and makes it available by default for all customers. No special setup or integration is required, and it's not accessible via an API—analysts access it directly within RiskOS™ Case Management. If you would like Socure to disable the feature, email [email protected].
In RiskOS™, analysts with permission to view a case can access explanations by clicking the Explain dropdown within each product tile—such as Fraud, Synthetic, or Email Risk—in the Case Management view.
Performance
A GenAI explanation has a latency of 5-10 seconds when it has not been viewed previously. Once an explanation has been generated and viewed, it loads within 1-2 seconds.
Sandbox availability
GenAI Explainability is currently not supported in the Sandbox environment, because Sandbox returns pre-configured model scores and underlying model features that are a prerequisite to generating explanations. To demonstrate the capability, Socure plans to provide pre-canned GenAI explanations in the Sandbox environment in the near future.
Troubleshooting
Sometimes, the AI explanation may not generate the first time you try it. If that happens, wait a few minutes and try again.
If you believe an AI explanation isn't helpful or is inaccurate, you can provide feedback for that specific explanation.
FAQs
General product behavior
Are the explanations unique for every evaluation?
Yes. Each explanation is generated based on the unique signals and data from that evaluation.
It's tailored to the actual score and contributing features, so no two explanations — much like evaluations — are exactly the same.
Is the response static or could it change if another user generates it later?
The explanation is static once it's generated.
It's saved to that evaluation and doesn't change — even if someone else clicks the button again.
This ensures consistency and auditability.
Will it still provide an explanation even if the score is low or not risky?
It can. If an analyst clicks the explanation button, the system still generates a summary — even for low-risk scores.
Is it explaining the specific risk signals (reason codes), or the model itself?
It explains the risk signals and features that most influenced the score — based on that evaluation's data.
It's not a technical deep dive into the model mechanics, but rather a plain-language summary of why the model responded the way it did.
Availability
Is this feature available to all new and existing customers by default?
Yes. This feature is available for all new and existing RiskOS™ customers.
In the rare case that a customer does not want this feature for compliance or other reasons, Socure's Support team can turn it off.
If a customer is using third-party data sources, will they get explanations too?
No. The explanations are generated for Socure's own products only.
Interactivity
Can users interact with the explanation — like asking follow-up questions?
Not yet. The current version provides a one-time, non-interactive explanation.
Interactive Q&A is on the roadmap for a future phase.
In later releases, users will be able to ask clarifying questions and get additional context.
Pricing
Does Socure charge for this feature?
No. The GenAI Explainability feature is free and available as a part of the RiskOS™ platform.
Updated 6 days ago